基于改进型RBF神经网络的磁流变阻尼器动力学建模及仿真
|
摘要
为提高磁流变阻尼器(MRD)动力学精度,提出一种网络连接权值自适应调整的改进型RBF神经网络模型。利用与任一测试样本相邻的两个训练样本对应的实际连接权值,对测试样本连接权值进行线性插值,提出连接权值的自适应算法;搭建MRD动力试验平台,进行多频率、多振幅的动力性能试验,利用大量实测力学特性数据,建立RBF神经网络模型以及连接权值自适应调整的改进型RBF神经网络模型,分析比较RBF神经网络模型在改进前后的平均累计相对误差变化规律,并进行数值仿真计算和试验测试分析。研究表明,在正弦激励频率0.25 Hz～1.0 Hz、振幅5 mm～15 mm、电流0～1.25 A工况下,相比于传统RBF神经网络模型5%的最大误差均值,改进型RBF神经网络模型使建模误差均值多控制在0.45%～0.85%之间,有效改善MRD的动力学特性,建模精度较好满足工程实际需要。
In order to improve the dynamic accuracy of magneto-rheological fluid shock absorber(MRD), an improved RBF neural network model with adaptive weights adjustment is proposed. Linear interpolation of test sample connection weights is performed by using the actual connection weights corresponding to two training samples adjacent to a test sample, the adaptive algorithm for the linear interpolation of the test samples is put forward, the dynamic test platform of MRD is built, the dynamic performance test of multi frequency and multi amplitude is carried out, and a large number of measured mechanical properties data are used to establish the RBF neural network model. The modified RBF neural network model with adaptive adjustment of connection weights is used to analyze and compare the average cumulative relative error changes of the RBF neural network model before and after the improvement, and carry out numerical simulation calculation and test analysis. The study shows that, under the condition of sinusoidal excitation frequency 0.25 Hz～1.0 Hz, amplitude5 mm～15 mm and current 0～1.25 A, compared to the maximum error mean of the traditional RBF neural network model 5%, the improved RBF neural network model makes the mean of modeling error control between0.45%～0.85%, effectively improving the dynamic mechanical properties of MRD, and the precision of modeling satisfies the project better.
In order to improve the dynamic accuracy of magneto-rheological fluid shock absorber(MRD), an improved RBF neural network model with adaptive weights adjustment is proposed. Linear interpolation of test sample connection weights is performed by using the actual connection weights corresponding to two training samples adjacent to a test sample, the adaptive algorithm for the linear interpolation of the test samples is put forward, the dynamic test platform of MRD is built, the dynamic performance test of multi frequency and multi amplitude is carried out, and a large number of measured mechanical properties data are used to establish the RBF neural network model. The modified RBF neural network model with adaptive adjustment of connection weights is used to analyze and compare the average cumulative relative error changes of the RBF neural network model before and after the improvement, and carry out numerical simulation calculation and test analysis. The study shows that, under the condition of sinusoidal excitation frequency 0.25 Hz～1.0 Hz, amplitude5 mm～15 mm and current 0～1.25 A, compared to the maximum error mean of the traditional RBF neural network model 5%, the improved RBF neural network model makes the mean of modeling error control between0.45%～0.85%, effectively improving the dynamic mechanical properties of MRD, and the precision of modeling satisfies the project better.
引文
[1]DENG Z H,ZHAGN D G,YAO X L.Experimental Study on Ship Equipments Vibration Reduction Based on Magnetorheological Damper[C]//Proceedings of the International Offshore and Polar Engineering Conference.2008.
[2]ANKITA S,DIPTESH D.Seismic Vibration Control of Offshore Jacket Platforms Using Decentralized Sliding Mode Algorithm[J].Ocean Engineering,2018,152(1):337-390.
[3]SHI D Y,SHI X J,REN L L.Experiment Research on Low-frequency Vibration of Ship Isolator[J].Journal of Vibration,Measurement and Diagnosis,2012,32(6):961-965.
[4]万乐坤,嵇春艳,尹群.海洋平台磁流变模糊半主动振动控制研究[J].船舶工程,2007,29(4):28-31.
[5]DYKE S J.Acceleration Feedback Control Strategies for Active and Semi-active Control systems:Modeling,Algorithm Development,and Experimental Verification[D].Indiana:University of Notre Dame,1996.
[6]HU G L,LIU Q J,LI G,et al.Simulation and Analysis of Adjustable Sigmoid Model for a Typical Magnetorheological Damper[J].Machine Tool&Hydraulics,2016,44(24):9-15.
[7]董致臻,冯志敏,伍广彬,等.基于粒子群优化算法的磁流变阻尼器多项式动力学建模方法[J].中国机械工程,2018,29(12):1421-1427.
[8]陈昭晖,倪一清.磁流变阻尼器非参数化模型泛化能力的提高[J].振动与冲击,2017,36(6):146-151.
[9]王檑,潘忠文,王旭,等.整星隔振用磁流变阻尼器神经网络模型[J].力学与实践,2017,39(6):579-584.
[10]AYALA H V H,COELHO L D S.Cascaded Evolutionary Algorithm for Nonlinear System Identification Based on Correlation Functions and Radial Basis Functions Neural Networks[J].Mechanical Systems and Signal Processing,2016,69(68):378-393.
[11]魏娟,杨恢先,谢海霞.基于免疫RBF神经网络的逆运动学求解[J].计算机工程,2010,36(22):192-194.
[12]李仁军,刘宏昭,原大宁.基于RBF神经网络的混合输入机构自适应控制[J].农业机械学报,2010,41(4):204-208.
[13]石为人,陶芬,张元涛.基于RBF神经网络的减摇鳍自适应滑模控制[J].控制工程,2012,19(6):978-981.
[14]冷志鹏,徐慧,李宏伟.一种新型磁流变阻尼器:中国,CN 102425636 A[P].2012-04-25
[15]冯志敏,张兴军,张刚,等.斜拉索-阻尼器系统建模与减振控制研究[J].农业机械学报,2013,44(S1):282-287.
[2]ANKITA S,DIPTESH D.Seismic Vibration Control of Offshore Jacket Platforms Using Decentralized Sliding Mode Algorithm[J].Ocean Engineering,2018,152(1):337-390.
[3]SHI D Y,SHI X J,REN L L.Experiment Research on Low-frequency Vibration of Ship Isolator[J].Journal of Vibration,Measurement and Diagnosis,2012,32(6):961-965.
[4]万乐坤,嵇春艳,尹群.海洋平台磁流变模糊半主动振动控制研究[J].船舶工程,2007,29(4):28-31.
[5]DYKE S J.Acceleration Feedback Control Strategies for Active and Semi-active Control systems:Modeling,Algorithm Development,and Experimental Verification[D].Indiana:University of Notre Dame,1996.
[6]HU G L,LIU Q J,LI G,et al.Simulation and Analysis of Adjustable Sigmoid Model for a Typical Magnetorheological Damper[J].Machine Tool&Hydraulics,2016,44(24):9-15.
[7]董致臻,冯志敏,伍广彬,等.基于粒子群优化算法的磁流变阻尼器多项式动力学建模方法[J].中国机械工程,2018,29(12):1421-1427.
[8]陈昭晖,倪一清.磁流变阻尼器非参数化模型泛化能力的提高[J].振动与冲击,2017,36(6):146-151.
[9]王檑,潘忠文,王旭,等.整星隔振用磁流变阻尼器神经网络模型[J].力学与实践,2017,39(6):579-584.
[10]AYALA H V H,COELHO L D S.Cascaded Evolutionary Algorithm for Nonlinear System Identification Based on Correlation Functions and Radial Basis Functions Neural Networks[J].Mechanical Systems and Signal Processing,2016,69(68):378-393.
[11]魏娟,杨恢先,谢海霞.基于免疫RBF神经网络的逆运动学求解[J].计算机工程,2010,36(22):192-194.
[12]李仁军,刘宏昭,原大宁.基于RBF神经网络的混合输入机构自适应控制[J].农业机械学报,2010,41(4):204-208.
[13]石为人,陶芬,张元涛.基于RBF神经网络的减摇鳍自适应滑模控制[J].控制工程,2012,19(6):978-981.
[14]冷志鹏,徐慧,李宏伟.一种新型磁流变阻尼器:中国,CN 102425636 A[P].2012-04-25
[15]冯志敏,张兴军,张刚,等.斜拉索-阻尼器系统建模与减振控制研究[J].农业机械学报,2013,44(S1):282-287.